import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
% matplotlib inline
import seaborn as snsimport warnings
warnings.filterwarnings('ignore')
# 不发出警告from bokeh.io import output_notebook
output_notebook()
# 导入notebook绘图模块
from bokeh.plotting import figure,show
from bokeh.models import ColumnDataSource
# 导入图表绘制、图标展示模块
# 导入ColumnDataSource模块import os

# 加载数据
data = pd.read_excel('双十一淘宝美妆数据.xlsx')data[data.isnull().values == True].drop_duplicates
# 检查缺失值
data.dtypes
# 查看数据类型

data.index = data['update_time']
data['date'] = data.index.day
# Series没有.day方法，利用index做了一步转换

data = data.drop('update_time', axis = 1)
data.reset_index(inplace = True)
data.rename(columns = {'店名': 'brand'}, inplace = True)data.head()

data1 = data[['id','title', 'brand', 'price', 'date']]items_num = len(data1['id'].unique())
brands_num = len(data1['brand'].unique())print('商品总数为: %d' %items_num)
print('品牌总数为: %d' %brands_num)

items_at11 = data1[data1['date'] == 11].drop_duplicates()
items_at11_num = len(items_at11)
precentage_11 = items_at11_num / items_numprint('双十一当天在售商品占比为: %.2f%%' %(precentage_11 * 100))

# 分析商品最早/最迟销售时间，及双11当天销售情况
data_sellrange = data1.groupby('id')['date'].agg(['min', 'max'])
id_at11 = data1[data1['date'] == 11]['id']data_temp1 = pd.DataFrame({'id': id_at11, 'on11sale': True})data1_id = pd.merge(data_sellrange, data_temp1, left_index = True, right_on = 'id', how = 'left')
data1_id.fillna(False, inplace = True)
data1_id['type'] = 'unclassified'
data1_id = data1_id.reset_index()
data1_id = data1_id.drop('index', 1)data1_id.head()# 分类
data1_id = data1_id.reset_index()
data1_id = data1_id.drop('index', 1)data1_id['type'][(data1_id['min'] < 11) & (data1_id['max'] > 11)] = 'A'
data1_id['type'][(data1_id['min'] < 11) & (data1_id['max'] == 11)] = 'B'
data1_id['type'][(data1_id['min'] == 11) & (data1_id['max'] > 11)] = 'C'
data1_id['type'][(data1_id['min'] == 11) & (data1_id['max'] == 11)] = 'D'
data1_id['type'][data1_id['on11sale'] == False] = 'F'
# 此时的F还包括E/H的情况
data1_id['type'][data1_id['max'] < 10] = 'E'
data1_id['type'][data1_id['min'] > 11] = 'G'
data1_id['type'][data1_id['max'] == 10] = 'H'data1_id[data1_id['type'] == 'unclassified']
# 检查分类是否完备# 分类的可视化
result1 = data1_id['type'].value_counts()
# 用于可视化的数据命名为result
result1 = result1.loc[['A', 'B', 'C', 'D', 'E', 'F', 'G', 'H']]
# 用于调整Series中Index的顺序

plt.rcParams['font.sans-serif']=['SimHei']
# 用来正常显示中文标签
plt.axis('equal')
# 保证长宽相等 --> 饼图为圆形
plt.pie(result1,labels = result1.index, autopct='%.2f%%',pctdistance=0.8,labeldistance =1.1,startangle=90, radius=1.5,counterclock=False)
# 绘制饼图

# 选择数据字段
id_not11 = data1_id[data1_id['on11sale'] == False]
df_not11 = id_not11[['id', 'type']]
# 此时没有title字段，需要找到data1合并
data1_not11 = pd.merge(df_not11, data1, how = 'left', on = 'id')
data1_not11.head()# case1 暂时下架
not11_case1 = data1_not11[data1_not11['type'] == 'F']['id'].drop_duplicates()
id_case1 = not11_case1.reset_index().drop('index', 1)
id_case1['case1'] = 1
# .unique()出现numpy数组，drop_duplicates()返回Series# case2 更名上架
not11_case2 = data1_not11.groupby(['id', 'title']).count().reset_index()
# 此时聚合的结果为每一个(id - title)对，对应的出现次数，即出现的日期和
title_count = not11_case2['id'].value_counts().reset_index()
# 计算id出现的次数，如果次数大于1，则说明商品更改过title
# reset_index() 使原来index里的信息成为一个新的列
id_case2 = title_count[title_count['id'] > 1][['index']]
id_case2.columns = ['id']
id_case2['case2'] = 1# case3 预售
not11_case3 = data1_not11[data1_not11['title'].str.contains('预售')]
id_case3 = not11_case3['id'].drop_duplicates().reset_index().drop('index', 1)
# .unique() 返回ndarray / .drop_duplicates() 返回Serise
id_case3['case3'] = 1# case4 未参与活动
not11_case4 = data1_not11[(data1_not11['type'] == 'E') | (data1_not11['type'] == 'G')]['id'].drop_duplicates()
id_case4 = not11_case4.reset_index().drop('index', 1)
id_case4['case4'] = 1print("未参与双十一当天活动的商品中:\n暂时下架商品的数量为%i个，更名上架商品的数量为\
%i个，预售商品的数量为%i个，未参与活动商品数量为%i个" %(len(id_case1), len(id_case2), len(id_case3), len(id_case4)))

# 查看没被分类的其他情况
id_not11_left = id_not11[['id', 'min', 'max', 'type']].reset_index()
id_not11_left = id_not11_left.drop('index', axis = 1)
#id_not11_left['case'] = ''
id_not11_left = pd.merge(id_not11_left, id_case1, how = 'left', on = 'id')
id_not11_left = pd.merge(id_not11_left, id_case2, how = 'left', on = 'id')
id_not11_left = pd.merge(id_not11_left, id_case3, how = 'left', on = 'id')
id_not11_left = pd.merge(id_not11_left, id_case4, how = 'left', on = 'id')
id_not11_left = id_not11_left.fillna(0)
id_not11_left['left'] = id_not11_left['case1'] + id_not11_left['case2'] + id_not11_left['case3'] + id_not11_left['case4']
id_not11_others = id_not11_left[id_not11_left['left'] == 0][['id', 'min', 'max', 'type']]id_not11_others.head()
# 未纳入分类的商品都是10日下架的，可能于双11当天售罄
# 暂时下架/更名上架/预售 可能互相包含

id_at11 = id_at11.reset_index().drop('index', 1)
id_presell = id_case3[['id']]
id_11all = pd.concat([id_at11, id_presell], ignore_index = True)
# 筛选双十一当日在售商品和预售商品id

id_presell['presell'] = True
id_11all['11all'] = Truedata1_id = pd.merge(data1_id, id_presell, how = 'left', on = 'id')
data1_id.fillna(False, inplace = True)
data1_id = pd.merge(data1_id, id_11all, how = 'left', on = 'id')
data1_id.fillna(False, inplace = True)data1_brand = data1[['id', 'brand']].drop_duplicates()
data1_id = pd.merge(data1_id, data1_brand, on = 'id', how = 'left')
# 合并数据 销售情况 id brand

data1_id.head()data1_on11sale = data1_id[data1_id['on11sale'] == True]
brand_on11sale = data1_on11sale.groupby('brand')['id'].count()data1_presell = data1_id[data1_id['presell'] == True]
brand_presell = data1_presell.groupby('brand')['id'].count()
# 统计各品牌双十一当天在售和预售商品数量# 利用bokeh绘制堆叠图
from bokeh.models import HoverTool
from bokeh.core.properties import valueresult2 = pd.DataFrame({'on11sale': brand_on11sale,'presell': brand_presell})
result2['11all'] = result2['on11sale'] + result2['presell']
result2.sort_values(by = '11all', inplace = True, ascending = False)lst_brand = result2.index.tolist()
lst_type = result2.columns.tolist()[:2]
colors = ["#718dbf" ,"#e84d60"]
# 设置好参数

result2.index.name = 'brand'
# 重命名标签
source1 = ColumnDataSource(result2)
# 创建数据

hover1 = HoverTool(tooltips = [('品牌', '@brand'),('双十一当天参与活动商品数量', '@on11sale'),('预售商品数量', '@presell'),('参与双十一活动商品总数', '@11all')])
# 设置标签显示内容
#output_file('折扣商品数量.html')

p1 = figure(x_range = lst_brand,plot_width = 900, plot_height = 350,title = '各个品牌参与双十一活动的商品数量分布',tools = [hover1, 'reset, xwheel_zoom, pan, crosshair'])
# 构建绘图空间

p1.vbar_stack(stackers = lst_type,x = 'brand', source = source1,width = 0.9, color = colors, alpha = 0.8,legend = [value(x) for x in lst_type],muted_color = 'black', muted_alpha = 0.2)
# 绘制堆叠图

p1.xgrid.grid_line_color = None
p1.axis.minor_tick_line_color = None
p1.outline_line_color = None
p1.legend.location = "top_right"
p1.legend.orientation = "horizontal"
p1.legend.click_policy="mute"
# 设置其他参数

show(p1)

data1_brands = result2.copy()
brand_total_item = data1_brand.groupby('brand').count()
brand_total_item.columns = ['total_items']
data1_brands = pd.merge(data1_brands, brand_total_item, left_index = True, right_index = True, how = 'outer')data1_brands.head()

Part2 价格折扣分析

1. 针对每个商品，分析其价格变化，判断是否打折
2. 针对在打折商品，计算其折扣率
3. 按照品牌分析，不同品牌的打折力度
4. 分析商家打折套路

bokeh_data4 = data_brands[['total_items', 'qzkl_mean', 'zk_pre']]
bokeh_data4['size'] = bokeh_data4['total_items'] * 0.2
bokeh_data4.columns = ['amount', 'zkl', 'pre', 'size']
bokeh_data4.reset_index(inplace = True)
# 筛选出品牌商品总数，打折商品比例，平均折扣率
# 将点的大小设置为与商品总数有关from bokeh.models.annotations import Span
from bokeh.models.annotations import Label
from bokeh.models.annotations import BoxAnnotation
# 导入注释相关模块

source4 = ColumnDataSource(bokeh_data4)
# 创建ColumnDataSource数据#output_file('各个品牌打折套路解析.html')

x_mean = bokeh_data4['pre'].mean()
y_mean = bokeh_data4['zkl'].mean()hover4 = HoverTool(tooltips = [('品牌','@brand'),('折扣率','@zkl'),('商品总数','@amount'),('参与打折商品比例','@pre')])
p4 = figure(plot_width = 800, plot_height = 800,title = '各个品牌打折套路解析',tools = [hover4,'box_select,reset,wheel_zoom,pan,crosshair'])
# 构建绘图空间

p4.circle_x(x = 'pre', y = 'zkl', source = source4, size = 'size',fill_color = 'red', line_color = 'black', fill_alpha =0.6,line_dash = [8,3])
p4.ygrid.grid_line_dash = [6,4]
p4.xgrid.grid_line_dash = [6,4]x = Span(location=x_mean, dimension='height', line_color='green',line_alpha = 0.7, line_width=1.5, line_dash = [6,4])
y = Span(location=y_mean, dimension='width', line_color='green',line_alpha = 0.7, line_width=1.5, line_dash = [6,4])
p4.add_layout(x)
p4.add_layout(y)
# 绘制辅助线

bg1 = BoxAnnotation(bottom=y_mean, right=x_mean,fill_alpha=0.1, fill_color='olive')
label1 = Label(x=0.0, y=0.60,text="少量大打折",text_font_size="10pt" )
p4.add_layout(bg1)
p4.add_layout(label1)
# 绘制第一象限

bg2 = BoxAnnotation(bottom=y_mean, left=x_mean,fill_alpha=0.1, fill_color='firebrick')
label2 = Label(x=0.35, y=0.60,text="大量大打折",text_font_size="10pt" )
p4.add_layout(bg2)
p4.add_layout(label2)
# 绘制第二象限

bg3 = BoxAnnotation(top=y_mean, right=x_mean,fill_alpha=0.1, fill_color='firebrick')
label3 = Label(x=0.0, y=0.80,text="少量少打折",text_font_size="10pt" )
p4.add_layout(bg3)
p4.add_layout(label3)
# 绘制第三象限

bg4 = BoxAnnotation(top=y_mean, left=x_mean,fill_alpha=0.1, fill_color='olive')
label4 = Label(x=0.35, y=0.80,text="大量少打折",text_font_size="10pt" )
p4.add_layout(bg4)
p4.add_layout(label4)
# 绘制第四象限

show(p4)

打折套路解析

• 少量大打折：雅诗兰黛、兰蔻、薇姿、悦诗风吟、欧珀莱
• 大量少打折：欧莱雅、玉兰油、美宝莲、妮维雅、蜜丝佛陀、美加净
• 大量大打折：自然堂、相宜本草、佰草集
• 不打折：SKII、倩碧、兰芝、娇兰、植村秀、资生堂、雅漾、雪花秀
• 假打折现象：妮维雅和美加净各有6件商品

Part3 汇总表格导出

• 商品id资料表格
• 品牌汇总资料表格
• 假打折商品详情

# 商品id资料表格
data_id_ch = data_id.copy()
data_id_ch.columns = ['最早上架', '最迟上架', '商品id', '双十一当天售卖', '销售时间分类',\'是否预售', '预售+双11当天', '品牌', '双十一前价格', '双十一当天价格',\'双十一之后价格', '前折扣率', '后折扣率', '假打折', '真打折']data_brands_ch = data_brands.copy()
data_brands_ch.columns = ['双十一当日商品数', '预售商品数', '双十一商品总数', '商品总数',\'真打折商品数', '假打折商品数', '双十一前均价', '双十一当天均价',\'双十一之后均价', '打折商品双十一前均价', '打折商品双十一均价',\'打折商品双十一之后均价', '平均前折扣率', '平均后折扣率',\'折扣商品占比']jdz_items_ch = jdz_items.copy()
jdz_items_ch.columns = ['最早上架', '最迟上架', '商品id', '双十一当天售卖', '销售时间分类',\'是否预售', '预售+双11当天', '品牌', '5日价格', '6日价格', '7日价格',\'8日价格', '9日价格', '10日价格', '11日价格', '假打折']data_id_ch.to_excel('商品id资料表格.xlsx')
data_brands_ch.to_excel('品牌汇总资料表格.xlsx')
jdz_items_ch.to_excel('假打折商品详情.xlsx')data_id.to_excel('data_id.xlsx')
data_brands.to_excel('data_brands.xlsx')
jdz_items.to_excel('jdz_items.xlsx')